Fact · Languages and accents
The Whisper authors say their worst outliers are languages with unique scripts that are distantly related to the Indo-European languages that make up most of the training data.
Checked against its source on October 17, 2026
In the source's own words
“languages that have unique scripts and are more distantly related to the Indo-European languages”
Robust Speech Recognition via Large-Scale Weak Supervision (full text), Alec Radford et al., OpenAI (arXiv), 2022, Section 3.3.
Open the original (DOI 10.48550/arXiv.2212.04356)- What it measures
- The paper lists Hebrew, Telugu, Chinese and Korean as examples. The authors suggest linguistic distance, a poor fit of the byte-level tokenizer, or data quality as possible causes.
- What it does not show
- These are the authors' suggested explanations, not tested causes.